Visualizations

Data Exploration

Data Preparation based on EDA

Model Building

Hyperparameters Tuning

Grid 1: Selecting class weight and estimators

Grid 2: Selecting max depth and split criterion

Checking if other depth and estimator value results better

Grid 3: Selecting minimum samples leaf and split

Grid 4: Selecting class weight

Final Model

save model

GAUGE

Final features

XGB

Feature Engineering

CUSTOMER SURVIVAL ANALYSIS

PREP

dummy variable creation

We need to drop variables such as customerID, tenure, Churn as they are not needed in survival data. Also, we need to add constant for survival analysis.

GENDER

Senior Citizen

PARTNER

dependent

PHONE SERVICE

MULTIPLE LINES

Internet Service

Online Security

Online Backup

Device Protection

TECH SUPPORT

STREAMING TV

STREAMING MOVIES

CONTRACT

PAYMENT METHOD

PAPERLESS BILLING

SURVIVAL REGRESSION

Survival Regression Ananlysis using Cox Proportional Hazard model

Saving the model

Customer Lifetime Value

To calculate customer lifetime value, I would multiply the Monthly charges the customer is paying to Telcom and the expected life time of the customer.

I utilize the survival function of a customer to calculate its expected life time. I would like to be little bit conservative and consider the customer is churned when the survival probability of him is 10%.